> Close to the Unix ideal of doing one thing, and doing it well.

In my opinion this concept does not work for most non-cli programs. How do you actually define this "one thing"?

Is "reading an email" one thing? And "writing an email" another thing? What about "spellchecking the email"? What about "managing all my email contacts"? What about "back up my emails"? What about "sign email with GPG"?

Do all of these things really need a separate program?

The unix principle works quite well for stuff like grep/find/awk but there is just now way you could make GIMP follow the unix principle.

> Is "reading an email" one thing?

That's essentially what Mutt does. But you could also use other programs to parse and read each email file if you wish. "Fetching email" can also be a different thing. I use offlineimap for that purpose and store my messages in the Maildir format.

> And "writing an email" another thing?

Yes. Even "sending an email" is another thing. The program that edits my emails (vim) is not the same program which sends them (msmtp).

> What about "spellchecking the email"?

I don't do this (with a program), but I assume you could just do it within Vim.

> What about "managing all my email contacts"?

I use OS X Contacts for that, so there's a command-line tool that allows me to search for and then insert those contacts in Mutt.

> What about "back up my emails"?

Again, another program! I keep my Maildir synchronized to my home server and all my other computers using BitTorrent Sync. Changes are backed up with Arq to Amazon Glacier every day.

> What about "sign email with GPG"?

Definitely its own program, although Mutt includes support for it you still need GPG installed and everything set up for that stuff in order to sign your emails. This is true for every other email client isn't it? I know I have to instal GPGMail to get the same functionality on Apple Mail.

So you see, all of these tools are really separate processes in and of themselves. The benefit to all of this is when newer and better technology comes around, you don't have to wait for your software vendor to support it. Instead, you just use a different program that follows the same standards, which have been around for decades. Some people, probably most, find this to be an unacceptable burden. For that, there are definitely great email clients out there like Nylas N1 and Google Inbox...but if you want total control in the same sense that you get from your shell and your editor, Mutt is the client for you.

Aspell will spell check, I know alpine will let you run this after the email is written, I don't know if mutt or pine will.

I was curious about your setup. Found this: https://gist.github.com/4540438

But no mention of OS X contacts integration. Care to share how you do it?

One way: https://github.com/tgray/contacts/blob/master/README.markdow...

Well, in case of mutt, it's just reading email from a local maildir. That is, a MUA (mail user agent) in it's own jargon.

Mostly everything else is outsourced. Composing is done by your $EDITOR, sending is done by an external program, fetching email by an external program too. Even indexing and querying is usually outsourced to a modern xapian-based tool, like mu or notmuch.

This leads to powerful composable workflows.

My own idea of the Unix ideal—I don't know how it accords with others'—is that there's nothing wrong with having an umbrella program that brings many pieces together, as long as those pieces are modular.

That is, I have no problem with being able to read, write, spellcheck, and sign an e-mail, manage my contacts, and back up my archive from within a single program; but it had better be true that, if I want to use vi instead of Nano to write my e-mails, then I can force the program to respect this wish.

(As is probably apparent, I'm a Pine user by experience; I've never used Mutt, but would like to give it a try.)

> The unix principle works quite well for stuff like grep/find/awk but there is just now way you could make GIMP follow the unix principle.

http://www.imagemagick.org/

;)

ImageMagick is a poor example of "doing one thing right" because every command line tool actually does a million different things with different command line switches. Additionally, it can't do everything GIMP can (which I think you're acknowledging with the winky face).

Imagemagick can't do everything GIMP can do. But everything Imagemagick can do can be scripted.

Need to convert 10, or 100, or 1,000, or 1,000,000 images? You might use GIMP for the first. You'd best give it up for the 2nd and further.

On Linux desktop, I had an Imagemagick command in my command history which would trim off the top n pixels of images to remove extraneous elements. Quick to call that up and run it on a new screenshot, or a series of shots, prior to posting online. The job was completed before GIMP would finish initialising.

"But everything Imagemagick can do can be scripted."

So can the Gimp. I wrote a calendar script in Perl for the Gimp literally 15 years ago.

My impression is that scripting Gimp is ... more complicated. That said, I've not tried it.

Imagemagick though makes simple-to-moderately-complex stuff bog easy.

better example would be the netpbm stack.

> Do all of these things really need a separate program?

Indeed. This is one of the reasons why the unixy "One app does one thing" philosophers dislike the C++ language so much. C++ is built for creating precisely the kind of apps you are describing - Classes are a well-defined compartment or component that does this "one thing" business while the app co-ordinates their working towards a common good.

C, on the other hand, pertains much to their "one thing" philosophy and they like it very much.

> Is "reading an email" one thing? And "writing an email" another thing? What about "spellchecking the email"? What about "managing all my email contacts"? What about "back up my emails"? What about "sign email with GPG"? > > Do all of these things really need a separate program?

There was a suite of programs that worked that way: https://en.wikipedia.org/wiki/MH_Message_Handling_System

Yup. I used it for a few years. It also had a Tk-based GUI. Every email message is a single file. "Folders" are directories. So you can use your usual text-processing commandline stuff: grep, sed, awk, etc.

http://rand-mh.sourceforge.net/book/mh/shomes.html#WeOBeYRe

Using MIME or GPG required a little configuration (mostly preset properly) which hooked external programs to preprocess before sending or displaying the message.

> The unix principle works quite well for stuff like grep/find/awk but there is just now way you could make GIMP follow the unix principle.

It seems that the unix principle is most relevant when compiling a C program takes too long and no interpreters exist that are fast enough. So, you figure out all the higher level functionality that you will need for system administration and compile that long before needing to do any sysadmin work. When a problem arises on a system, you can write a shell script that uses those already compiled commands, instead of writing a C program and waiting for it to compile. If your initial code sucks, you don't have to debug and recompile; you just debug and re-run the script.

Compiling C programs is now substantially less time consuming, so you could use C for sysadmin work (shudder). C shell was perhaps a similar attempt to get the benefits of C, but without the downside of compilation times. I seem to recall some essay on the harmfulness of using csh, so C will continue to keep its place.

how about we break down each action one can take in the GUI as its own binary that is part of a larger set? that way the GUI becomes a wrapper that do one thing, wrap the binaries that all their own thing, but that gets composed to do larger jobs.

One IMO important thing that's missing from others' replies: It's extremely simple to integrate other programs for processing your mails. Want to be able to import patches sent to you into some git repository? Just write a small shell script that accepts the mail (or some attachment) as input and then create a keyboard shortcut in mutt that pipes the mail (or attachment) to that script, done. You don't need to extend mutt in order to be able to programmatically process your emails, you simply use standard unix mechanisms (a pipe or maybe a file) and call an external program.

> In my opinion this concept does not work for most non-cli programs. How do you actually define this "one thing"?

I don't see how CLI programs avoid this. "One thing" is a human idea, subject to interpretation. Should grep and sed be consolidated? Does find do too much? Why do we have both echo and printf?

> Is "reading an email" one thing? And "writing an email" another thing?

Mutt says so. You read and sort emails in Mutt, you write emails in $EDITOR and then you can add attachments and such back in mutt.

> What about "spellchecking the email"?

Well since mutt outsources writing the email, I can't really comment on what mutt does. But if you don't want to do spellchecking within your editor for whatever reason, mutt lets you pipe the contents of anything to an external program, so you could write your email, then select the message from the attachments list and type "|aspell -a<CR>" and you'll get a basic spellcheck.

> What about "managing all my email contacts"?

I think that's pretty clearly a seperate program's duty, since contacts are used by other applications as well. Not only does mutt agree with me, so do Google and Apple. Mutt makes it simple to integrate with external contact lists to provide autocomplete for known email addresses.

> What about "back up my emails"?

That should be taken care of by both your email provider and your personal backup solution. I don't see why an email client would care about backups.

> What about "sign email with GPG"?

Well that's a job for GPG isn't it? That again can be accomplished by just piping the message into an external command. If you use it frequently you'd write some config and scripts to automate it. I saddly rarely uses PGP for email so I don't have hands on experience with this, but it looks like mutt has built in support for this kind of automation. A more unix-aligned approach would be to ship an /usr/share/mutt/examples/auto-gpg.muttrc or something like that that users can just source in their config files.

> but there is just now way you could make GIMP follow the unix principle.

No? Represent layers as a bunch of image files in a directory. Have one program check when they change and composite them in realtime. Have another program render the output of that one to the screen. Have a third integrate with the rendering program and manage the creation of selections (represented as a selection mask, so it supports everything from rectangular selection to brush selection). Drawing tools and filters become programs that take an image and a selection and some (interactive) options and output a new image. Then you have a UI program that brings all these programs together.

If this isn't as performant as Gimp, you can optimize it, for example, by replacing pipelines between parts of the standard toolchain with some kind of shared memory IPC or something similar.

It's not that the ideas behind unix don't apply to modern software; people just don't give a fuck.

Hats off for the gimp-architecture review

Being just a little bit hard-arsed about these things can pay off dramatically.

> Is "reading an email" one thing?

In fact, sometimes it can be the "only" thing you want. Most often it is the case with open source developers when they want to catch up a huge discussion thread. (Like bugs to email interface). Mutt is excellent here.

If I have something to add to the discussion, I have the option to use the command line or log in to web interface and post my response.

See (n)mh:

http://rand-mh.sourceforge.net/book/mh/shomes.html

Have you used the app yourself? Could you maybe talk more about what you did not like about it?

I did and liked it very much. Still the person you are responding to has a valid point IMO.

I personally always felt that "full applications" were disconnected from the Unix philosophy. By definition, such an application does many things: fetch things from the network, display them, sort them, ... it talks many protocols, concerns itself with many I/O systems.

I'm certainly not saying that it is the way to go, but a real Unix philosophy way of dealing with email would be to use a command to pull mails from a mail server as text files (heck, as a text stream!), then other commands to display them, sort them, ...

Well, mutt is close to what you say. It's just a MUA [1].

That is, it just reads emails on a local mailfolder. Fetching emails to that mailfolder from the server, sending emails and doing other operations is a task of 3rd party programs.

[1] OK, it has a few addons for lazy people. But most experienced users ignore them.

That's what mutt is. You can compile it with support for imap and smtp, and most packaged versions of it are built that way, but the traditional way of using it is just as way of reading and sorting emails. Fetching emails from the server is done by a POP or IMAP agent (I use offlineimap), sending mail is handled by an smtp agent (I use msmtp), composing emails is done with $EDITOR.

I don't think that mutt is the most unix-philosophy-aligned MUA possible, it feels kind of like GNU tools: unixy, but jam packed with too many flags and options and convenience features. The unix philosophy would prefer that the application's design be changed to better accomodate implementing those extra features externally where plausible. But the perfect unix MUA would do the same things as mutt.

Actually, the traditional way (that still works perfectly fine, actually) is to not "fetch emails" at all. You simply have an SMTP server on your machine and have it either be the MX for your domain or have the MX forward emails for your address(es) to your machine via SMTP, and then have the MTA deliver to the mailbox file or directory that Mutt reads from.

I think a lot of Mutt users do what you are describing.

I use mbsync to synchronize my mail over IMAP with my remote mailserver. I use mu to index it for search. I use msmtp to send mail. And I use mutt to read and write mail.

I didn't know, thank you everyone!

I used to be a mutt user back in the day, and it was great at the time, especially when unix/linux lacked anything really better, and there were no iphones or android devices. It was/is the best-in-class console based email client, probably still is.

I can't but help feeling though these days it just doesn't really gain you anything substantive to bother to learn it, or even continue to use it, and spend the hours it will require to dial in and maintain your configuration.

Most gui clients are good enough.. webmail (Gmail and alternatives) are good enough, unless you have some non-negotiable requirement that mandates you have your own self-hosted email servers, I have a hard time believing its not an overall productivity loss to spend time in mutt (or Gnus for emacs, and other such things).

If you like to tinker with that kind of thing, and you like making your email client a hobby of sorts, cool. But what do you really gain other than a techno-bauble to tinker with?

"and spend the hours it will require to dial in and maintain your configuration."

This is my main gripe with mutt. The days where I could futz around with 100 config setting for hours are long behind me. Just having to set up an smtp daemon every time, and having to maintain it, is such a pain in the ass. Probably most tools have moved on by now (I moved away from mutt in 2004 or 2005) but I used mostly postfix and qmail (when I got fed up with one I switched to the other for a few months, etc). Just having to remember the dozens of idiosyncrasies in those was enough of a PITA to make me switch to an integrated graphical mail program.

My main method of searching email was grep -R ~/.Maildir . When I think back about that - ugh.

(n)mh is the Unix ideal.

Once you are "inside" an app instead of your shell prompt, you are no conforming the Unix ideal.

the UNIX philosophy died with the advent of the web imo. The emacs editor/Mozilla suite (SeaMonkey) mentality of doing most things and doing them relatively well is the reality we have inherited.

Not the case in my reality.

Emacs has always been That Way. And then CERN/Mosaic happened, and browsers (being all-singing, all-dancing environments that do everything from page layout to hosting interpreters and databases, are also big apps) came along.

This has not lead to awk becoming an MUA, or tmux adding a flight simulator, or tee offering to store things as PDFs. Nor has it stopped the development of building-block tools - I still see more nifty new things built with more or less the same assumptions than I can play with.

The only thing I can think of is that you're referring to GUI apps, at least mostly on Linux. GUIs (or more specifically, what Alan Cooper termed 'sovereign applications', but they are mostly GUIs) are typically subject to Zawinski's Law in ways that command line tools are not. Heck, Mutt is somewhat subject to it, itself, for that reason.

Like most pop-philosophy, the Unix philosophy is too vague and subject to interpretation to really be a good guide or yardstick for your projects or ideas.

Its a nice little useful heuristic or reminder to ask yourself if maybe what you are doing is too much.

But maybe what you are doing isn't too much.

That's all it should ever be, IMHO... a reminder to check yourself at some point or another - but its perfectly valid to conclude that your monolithic, large app can and does many things well, and cohesively adds something of value to the software world, that can't or hasn't been accomplished by tiny composable bits, or at least not nearly as well.

The UNIX philosophy is still alive and kicking and just a terminal away courtesy of SDF.org.

Love SDF. (Of course, I run bash and Emacs on it, which will probably earn me a slightly-warmer spot in Hell someday than if I used ksh and vi like God intended.)

Sacrilege. Ksh93 and nvi here.

<Hoping no-one runs "w" while I'm in elinks>

It was redone poorly as `npm install reinvented-wheel`.

As someone said (paraphrasing): Emacs is a great operating system, but its text editor could be a little better.

>Nothing to envy from Gmail.

What do you use for mobile access?

I use k-9 mail these days, which is OK but doesn't handle huge folders well. However, back when I had an n900 with its hardware keyboard, my mobile mail client was ssh, screen, and mutt.

you can have that again: bbm priv.

An android phablet with a full hardware keyboard.

Or a Pyra!

That thing is CHUNKY.

Some decent pictures: http://pandoralive.info/?p=5519

Already possible with an Open Pandora :)

not the commenter you want, but i had both a mobile client (if i am using mutt, that means i can access the imap/pop server from other clients) and the ultimate ssh client on my phone which gave me full mutt, in case i needed to have my procmail filters helping me)

Navigating any terminal software without a true keyboard is an exercise in frustration. Look, I understand what you're saying, but claiming that an ssh client on a phone is a full-fledged replacement for a touch-optimized UI is disingenuous.

> Nothing to envy from Gmail.

I switched because of search, which is amazing in gmail and hard in mutt (for me). What do you do?

> search amazing in gmail.

really? until a few months ago, you couldn't find anything besides whole words with textual search on gmail. why do people have it's search at such high standards? how bad was the alternative you had?

It's funny, i have been frustrated since day 1 which gmails search. Granted, it could very likely be user error - as i often forget the set of keywords that might lead me to the email i desire, but regardless i as a user am left frustrated.

Since someone may ask, my desires are simple in nature, but i'm sure difficult to implement. I simply want to find an email. On the Google Search side of things, i have often been very impressed by the search results i am able to achieve. Yet for email, it's been far less successful.

On Search, i'm able to use synonyms, or hell even vaguely describe (in the case of movies) plots, and often be presented with what i want. On Gmail, it feels like Gmail has no idea about the context or content of my mail (which is arguably their selling point), and instead i'm purely searching by keyword. The frustration is amplified by the fact that Search(proper) works so surprisingly well, i'm sure.

Search has the benefit of all they data they have from the millions of other users who searched the same term and clicked certain results. It also has the backing of Pagerank. Your email is unique to you and doesn't have any of these advantages.

What are you talking about? What have you've been missing?

Senders name or email? That has always been an option. Has attachments? Been an option since forever. tags? Since forever. date? Available since forever.

Not OP, but here's something I miss in Gmail search all the time-- ability to target a search at a particular directory and its sub-directories.

Suppose, for example, I have set up a subdir for conferences. That subdir contains a bunch of subdirectories for individual conferences. I know the e-mail I'm looking for is stored in there, because it relates to a conference publication I wrote. However, that subdir itself is "empty" as far as gmail search is concerned (because it's a "tag" rather than a true subdirectory).

The net effect of these hollow imitations of directories is I can either target the search to a specific tag, or I can search throughout my entire collection of messages. (Or, I suppose, I can tag every e-mail with every single sub-tag represented in the directory hierarchy I have mentally set up, just in case I might want to search for it someday.)

The gmail ideal, of course, is that I abandon the notion of directories altogether, and just let all my messages stew in the ocean of e-mail that one accumulates over the years. Supposedly their "powerful" search capabilities will help me find what I want. In my experience, though, under this approach most reasonable searches produce multiple pages of possible matches. This, of course, is hardly efficient.

It would also be nice if I could perform a search and then pull out that page of returned results into a separate screen/window for ongoing reference. Often when I'm searching for something, I'm looking broadly for resources or related conversations, rather than for a specific message.

Another place where I feel gmail search is brain-dead is its lack of regex capability (which @devdas has already mentioned).

Check again - gmail supports nested labels. exactly same as directory/subdirectories.

Separate screen of search results is just another tab in your browser - that's actually a vote in favor of web based gmail.

Yes, it supports nested tags, and the tags create structure that suggests directory nesting. However, as far as I can tell the search can't be made to honor the nesting.

To make my example explicit above, suppose I have a "directory" structure with dirs like `conf`, `conf/iaq2016`, and `conf/gmu2012`. Then a search in `conf` will not turn up results in `conf/iaq2016` or `conf/gmu2012`.

Believe me, I check this once every few months, always hoping that google will have improved their "tag" implementation.

I do appreciate your tip about multiple browser windows, though. In general, I don't like having multiple windows open to the same application-- but that's my problem, and not a design flaw in Gmail's search.

Yup. Nested labels like "top/med/bot" get flattened to "top-med-bot" for search, e.g. in:top-med-bot keyword

Regexp.

RIP Google Code Search :(

For reference, you used to be able to do this - https://code.google.com/p/chromium/codesearch - with an archived cross-section of code on the Web.

I wonder if we'll ever be able to do regex Google searches in the future, or if AI/machine learning will make that sort of flexibility redundant by the time the hardware would be able to handle it.

FWIW codesearch's spirit lives on with debian code search (https://codesearch.debian.net/) which allows you to do the same as codesearch, except for Debian only. Checkout also the thesis behind (https://codesearch.debian.net/research/bsc-thesis.pdf) that explains a lot about the challenges and the architecture.

It's cool to hate on Google though I thought?

But seriously, all these hard core unix fans seem a little odd here. Gmail search has always been powerful, and is even more powerful now. It sounds like a case of user error if it isn't working for someone.

I'm not sure what you mean here - I've searched in email for years with gmail - and it's basically always been flawless. Search on From:, To:, Subject, Date, etc... - in addition to searching for content in the body.

I'm curious, what other search does one do in email?

The search I have the most trouble with is searching for things that match a label and are in my inbox. GMail will consistently miss messages. I don't know what it is about them, but I imagine it's some indexing bug. I can go to my inbox, click through the first several pages of messages, and see that a message is there, and has my label. I can go to my label, and find it there too. But a search for 'label:foo in:inbox' will come up empty.

I use notmuch. It's really powerful and quick. You get all Gmail features, plus some more.

There's also mutt-kz, with notmuch deeply integrated.

I use `notmuch` for searching local mail. It's pretty powerful and blazing fast.

I use mairix, which is very old but still does the job.

grepmail

zcat Mail/2015/mbox.gz | grepmail -i whatever | grepmail 'perl.style.*regex' >myresults

mutt -f myresults

I wish I could interface NNTP and RSS feeds with Mutt somehow.

I've been using rss2email (https://github.com/wking/rss2email) to read rss feeds in mutt. It works well for me.

I actually use this already.

I think my dissatisfaction with mutt + rss2email is more about RSS in general. It amounts to just being a short description of the article and a hyperlink, and at that point I might as well just load these websites in my browser.

I need a better article scraper. General solutions that try to detect content, like Firefox's article mode and similar, don't do enough.

We need youtube-dl for text.

Maybe a little more heavy handed but I use huginn for exactly that. Read in numerous RSS feeds and digest them into my email box. It does a lot more than that too.

https://github.com/cantino/huginn

Wow I can't believe I've never heard of this. This is cool. Thanks for posting it.

Shameless plug: I use sinntp[0] to grab newsgroup messages and put them into a mailbox. You can also configure mutt so that it can post to newsgroups (see doc/mutt-integration.txt).

[0] http://jwilk.net/software/sinntp

This looks like good software. I'm going to try using this to replace slrn. If it's okay with you, I'll email you in a couple weeks with an update.

Alpine has NNTP support.

Not quite what you want, but in the same spirit of Mutt is Newsbeuter.

http://newsbeuter.org/

You could use rss2email on a cron-job to mail you RSS things - it's not as pretty as a console reader, but it might help?